The effect of KIR-blocking on HLA-C immune response to SarsCov2

Emilie Sofie Engdal, Manon Kirstine Chevalier

Introduction

  • Unusually high number of HLA-C recognition in CD8+ T cells

  • Responses could not be validated through TCR-specific confirmation assays

  • The Killer Immunoglobin Receptor (KIR) leads to false positive responses

Figure 1: KIR binding and blocking (Source: Biorender)

Materials and Methods - Project Flow Chart

Figure 2: Flowchart of the project

Materials and Methods

Data:

  • High-throughput experiment

  • Thousands of SarsCov2 peptides binding to T-cells

  • TCR-dependent or TCR-independent (i.e.: KIR-mediated)?

3 cohorts:

  • AP: Leukemia patients who have had COVID-19 (n = 12)

  • BS: Healthy nurses who have had COVID-19 (n = 17)

  • BC: Healthy donors (n = 15)

Methods:

  • DNA-barcoding: Tagging each peptide-HLA pair uniquely

  • Flow cytometry: Sorting T-cells based on TCR vs KIR binding

Materials and Methods - Barracoda data

AP
(N=13465)
BS
(N=7840)
BC
(N=10491)
Overall
(N=31796)
KIR_blocking
1 7541 (56.0%) 4019 (51.3%) 7877 (75.1%) 19437 (61.1%)
0 5924 (44.0%) 3821 (48.7%) 2614 (24.9%) 12359 (38.9%)
log_fold_change
Mean (SD) -1.15 (3.29) -2.25 (3.95) -1.49 (3.69) -1.53 (3.62)
Median [Min, Max] -0.182 [-13.4, 9.71] -0.508 [-12.0, 8.64] -0.313 [-12.4, 11.2] -0.290 [-13.4, 11.2]
protein
ORF1 7239 (53.8%) 4447 (56.7%) 6103 (58.2%) 17789 (55.9%)
Nucleocapsid 707 (5.3%) 434 (5.5%) 548 (5.2%) 1689 (5.3%)
Spike 2619 (19.5%) 1581 (20.2%) 2062 (19.7%) 6262 (19.7%)
ORF6 68 (0.5%) 42 (0.5%) 65 (0.6%) 175 (0.6%)
Membrane 785 (5.8%) 465 (5.9%) 545 (5.2%) 1795 (5.6%)
ORF3 540 (4.0%) 345 (4.4%) 467 (4.5%) 1352 (4.3%)
ORF7 288 (2.1%) 175 (2.2%) 227 (2.2%) 690 (2.2%)
Envelope 119 (0.9%) 73 (0.9%) 91 (0.9%) 283 (0.9%)
ORF10 119 (0.9%) 73 (0.9%) 91 (0.9%) 283 (0.9%)
ORF8 322 (2.4%) 205 (2.6%) 292 (2.8%) 819 (2.6%)
Unknown 659 (4.9%) 0 (0%) 0 (0%) 659 (2.1%)
HLA
HLA-C04:01 300 (2.2%) 600 (7.7%) 2400 (22.9%) 3300 (10.4%)
HLA-C06:02 2695 (20.0%) 1086 (13.9%) 1164 (11.1%) 4945 (15.6%)
HLA-C07:01 4014 (29.8%) 2950 (37.6%) 2655 (25.3%) 9619 (30.3%)
HLA-C07:02 6456 (47.9%) 3204 (40.9%) 4272 (40.7%) 13932 (43.8%)

Table 1: Barracoda data

Figure 3: Protein representation accross the 4 HLA-C subtypes

~ 32 000 bar-codes (patient-HLA-peptide combinations)

  • 44 patients

  • 4 HLA types

  • 11 protein types

Results - Filtering only significant responses

The number of observations reduced drastically during wrangling:

Figure 4: Volcano plot of all responses in joined dataset

  • Original: 31796

  • w. Flowjo: 22819

  • Significant: 542

Results - Paired boxblot

Paired data to compare TCR-dependent and TCR-independent responses:

Figure 5: Effect of KIR blocking

This further reduced the dataset size:

  • Original: 31796

  • Significant: 542

  • Paired: 22

    • 11 pairs

      • 1 normal

      • 1 KIR-blocked

Results - Bubble plot - paired

\[ Est. Freq = \left( \frac{count_{barcode}}{count_{total}} \cdot 100 \% \right) \cdot PE \]

Figure 6: Responses across pMHC multimers in the paired dataset

  • The fraction of T cells recognising the specified peptide

  • C0602 is a true response

  • AP cohort may have lower KIR expression due to pathology

Results - Bubble plot - unpaired

Figure 7: Responses across pMHC multimers in the unpaired dataset

  • BC cohort show significant response for C0401

  • False positives are removed in BS cohort for C0602

  • C0701 and C0702 show inconsistent responses

Discussion

  • KIR- blocking was showed to reduce the binding of pHMC
    • corroborates TRC-independent binding
  • KIR-blocking was less efficient in reducing the binding of pHMC in the AP cohort
    • hematological patients may have a lower KIR expression
  • Some results suggest that KIR-blocking is not as robust/reliable as expected

Limitations

  • The paired dataset was really small, hard to draw reliable conclusions

  • New info from this research group indicates that the KIR-antibody used in this experiment only blocked KIR-C1 but not -C2

    • C0401 and C0601 are C2 ligands and were therefore not blocked

    • This would explain the unreliable results, and variation between cohorts